Understand how your sounds are perceived.
A structured listening study with independent listeners, designed to compare sound variants and show where the evidence is clear — and where it isn't.
Everyone on the team hears something different.
One person says variant A feels calmer. Another says B is more urgent, in a good way. Nobody's wrong — perception is exactly that personal. But a product decision needs more than confident opinions in a meeting room. It needs evidence from people who don't know your product, your roadmap, or which version you prefer.
A structured process from question to evidence.
DoReMiFo runs this pipeline for its own ongoing perceptual research — a live testing environment, an active participant pool, and an evaluation methodology already in use.
You send your sounds and the decision
What you're deciding, and for whom.
DoReMiFo designs the experiment
Using an evaluation methodology already in use in ongoing research.
Independent listeners are recruited
Through an existing, active Prolific account, screened for attention and consistency.
The listening test runs
On DoReMiFo's own testing environment, already proven at scale, able to test dozens of sounds per round.
The data is analyzed
AI-assisted statistical modeling — see "Where AI is — and isn't" below.
You receive the evidence
How each sound lands, the variant comparison, Decision Confidence — and one of three calls: Ship, Revise, or Evidence inconclusive.
A clear choice when the evidence supports one.
A clear limit when it doesn't.
- How each sound lands — tone, energy and the meaning listeners read into it
- Direct comparison between variants
- Intended meaning: clearly recognised or ambiguous — how each sound reads against the event it has to signal
- Decision Confidence — the strength of evidence behind the call
- One of three calls: Ship, Revise, or Evidence inconclusive
AI-assisted, not AI-decided.
Analysis is AI-assisted: statistical models turn listener responses into the profile, comparison and confidence estimates in your report. The listening itself isn't — every rating comes from a real, independently recruited listener, not a prediction.
ICC measures how stable a sound's average rating stays when you draw a different group of listeners — not how much two individual people agree with each other. It's a property of the panel-level measurement, not a claim that any two listeners rate a sound identically.
A person on the DoReMiFo team reviews the analysis and interpretation before it reaches you. AI assists the process, it doesn't sign off on your result.
Built on the same methodology as the research.
DoReMiFo is a research lab: commercial listening studies (DoReMiFo s.r.o.) and academic research (HAMU Prague, SGS grant) inform each other while remaining separately funded and governed.
Using client data for further research is never automatic — it depends on the agreement in place for that engagement.
Rigorous about what we know.
Explicit about what we don't.
If the evidence does not support a meaningful difference, we say so.
HAMU Prague
SGS Academic Grant
Russell Circumplex (1980)
Essentia · active-segment
Two "error" sounds. Same job. Very different profile.
From DoReMiFo's own calibration measurements: two everyday error sounds you already know, rated by independent listeners on emotional tone (positive–negative) and energy (calm–high-energy). Anonymized here — the point is the gap between them, not the brand.
Two sounds doing the exact same job — telling you something went wrong — can land in very different places. That gap is exactly what a Sound Perception Test is built to catch before you ship, not guess at after.
Built for the sounds that shape everyday interaction.
Product and UI sounds
The small sounds that confirm, deny, or guide an action.
Notifications, confirmations and alerts
Moments that need the right urgency — not just more of it.
Game sound effects
Feedback that shapes how an action feels, round after round.
HMI and device sounds
Interface sound for physical products, appliances and vehicles.
Short sonic branding assets
A few seconds of sound that carry an identity.
Questions worth answering upfront.
How long does a test take?
A few days from your brief — send your sounds and the decision you're making, and we'll confirm the exact timeline before you commit.
How many listeners are involved?
Our target is 25 independent listener ratings per sound — enough for a statistically meaningful comparison, consistent with DoReMiFo's established methodology.
Is my data used for future research?
Not automatically — it depends on the agreement in place for your engagement.
What if listeners don't agree with each other?
That's exactly what the test is designed to surface. Your report includes the strength and consistency of the evidence, not just an average — if listeners are split, we say so instead of picking a winner for you.
Do you compose or generate sounds?
Not within the Sound Perception Test — this service measures and evaluates sounds you already have. It doesn't compose or generate audio for you.
Not sure whether a listening test fits your decision?
You do not need to prepare a brief or upload sounds. A short description of what your team is comparing is enough.
Email Bohdan